Charles has been working With Dr. Akinsete in the department of Mathematics. This summer, he was working on Stochastic modeling of sports data.
"The first part to my project was to review various papers on previous work on soccer. After becoming acquainted with past works, we began to analysis data from the 1995/1996 to 2005/2006 seasons of the FA Premier League of England. I first did a table showing the times that one team had played another and had either won, drawn, or lost. Later we would calculate probabilities from this table in the same file. Next was the creation of a table showing the movement of all of the teams in the last three years from round to round of play. Then a table showing the transitional probability of moving from win to draw to loss was made. A table was created using the "run test" to show the randomness of each team. Another table was created to show the transition of teams from year to year during the 11 years. The home ground advantage theory was proved by another table and t-test. The final table made shows the Clark and Norman model applied to the last three years of our data. With all of the data collected for our research, we are now finishing the analysis of three of the tables using the (P=( equation and beginning to write the paper suitable for publication."
Progress: Stochastic Modeling of Sports Data -
A Case Study of English FA League Soccer Results
The work investigates the stochastic behavior of the FA Premier Football League results from 1995/1996 to 2005/2006. The initial stage of the project consists of the review of many relevant papers on previous statistical analysis of soccer data, with particular reference to the model proposed by Akinsete (2004). The author proposed a ranking model of teams and obtained the winning probabilities of the teams in the league. The review of methodologies includes among others, the reading of a book on stochastic processes, a course taught at the graduate level. Following this was the capturing, verification and validation of the data, consisting of 4,180 data values. The following results were obtained:
- Probabilities of teams losing or tying with another in the league.
- The transitional probabilities of moving from the winning state to another.
- Investigation of the run test, showing the randomness in the performance pattern of the teams.
- The year to year transition or placement of teams at the end of the league seasons.
- Investigation of the home (ground) advantage, using various descriptive measures.
- The computation of the Clark and Norman (1995) model, applied to the last three years of our data.
- Steady state or long run performance probabilities of teams in the league.
A time period is needed, separate from the 10-week research period in summer, for the compilation of these results. The results are intended for publication and presentation at a national conference in summer 2007.